For example,Бобцов

COMPARISON OF UNKNOWN PARAMETERS ESTIMATES BY METHOD OF DYNAMIC REGRESSOR EXTENSION AND MIXING AND LEAST SQUARE METHOD IN NOISE PRESENCE

Annotation

Subject of Research. The paper deals with identification of the regression model unknown parameters by two estimation algorithms: the classical least square method and a new method of dynamic regressor extension and mixing. To compare the quality of the obtained estimates, the non-stationary parameters of the regression model were considered, and various noise of limited power was added to the input signal of the model. Method. The problem was solved by the dynamic regressor extension and mixing method followed by the gradient algorithm and the least squares method in real time mode ignoring the older values of the measured input signal. Main Results. The numerical simulation was presented illustrating the qualitative comparison of the two used methods. A noisy shifted sinusoidal signal with time-varying and unknown parameters of displacement, amplitude and phase shift was applied to the input of the estimation algorithms. Comparison has shown that with the application of dynamic regressor extension and mixing method, the estimation of the input signal parameters had an aperiodic form, while the least squares method gave unwanted oscillations. Numerical simulation has shown that the method of dynamic regressor extension and mixing qualitatively exceeds the method of least squares. Practical Relevance. The results can be used when solving practical problems in the areas of processing and evaluating for both harmonic signals and also signals with a more complex form.

Keywords

Постоянный URL

Articles in current issue

CONTENT © ITMO UNIVERSITY 1993-2024